Dynamic weighted ensemble model for predictive optimization in green sand casting: Advancing industry 4.0...

Area of Science:

• Materials Science and Engineering
• Manufacturing Technology
• Data Science and Machine Learning

Background:

• Casting defects significantly impact product quality and increase rejection rates in manufacturing.
• Existing predictive models struggle with the nonlinear complexities of interdependent process parameters in green sand casting.
• Limitations of individual machine learning models and static ensemble methods necessitate advanced predictive strategies.

Purpose of the Study:

• To introduce a novel Dynamic Weighted Ensemble (DWE) model for enhanced defect prediction in green sand casting.
• To address the nonlinear complexities and interdependent parameters inherent in the casting process.
• To improve the accuracy and consistency of defect prediction beyond individual machine learning models.

Main Methods:

• Developed a Dynamic Weighted Ensemble (DWE) model that adaptively assigns weights to algorithms based on performance.
• Evaluated five machine learning models (Linear Regression, Ridge Regression, Decision Tree, Random Forest, Gradient Boosting) using 10-fold cross-validation.
• Selected the top three performing models based on Root Mean Square Error (RMSE) for ensemble integration.
• Applied the DWE model to five-fold unseen test data for performance evaluation.

Main Results:

• The DWE model achieved an average RMSE of 8.07 on unseen test data.
• Demonstrated a 2.1% improvement in RMSE and a 2.3% increase in prediction accuracy compared to the best individual model.
• Statistical analysis (paired t-test) confirmed the significant improvement (p < 0.05) and superior prediction consistency of the DWE model.
• Identified Gradient Boosting (RMSE: 8.25), Ridge Regression (RMSE: 8.30), and Linear Regression (RMSE: 8.31) as top individual performers.

Conclusions:

• The proposed DWE model offers a robust and adaptive solution for defect prediction in green sand casting.
• The enhanced prediction accuracy and consistency support real-time optimization and quality improvement in manufacturing.
• The DWE model aligns with Industry 4.0 principles by facilitating automated, data-driven decision-making in smart manufacturing environments.